Download Impacts of climate change on temperature and evaporation from a

Griffith Research Online
https://research-repository.griffith.edu.au
Impacts of climate change on
temperature and evaporation from a
large reservoir in Australia
Author
Helfer, Fernanda, Lemckert, Charles, Zhang, Hong
Published
2012
Journal Title
Journal of Hydrology
DOI
https://doi.org/10.1016/j.jhydrol.2012.10.008
Copyright Statement
Copyright 2012 Elsevier B.V.. This is the author-manuscript version of this paper. Reproduced in
accordance with the copyright policy of the publisher. Please refer to the journal's website for access to
the definitive, published version.
Downloaded from
http://hdl.handle.net/10072/49475
Impacts of climate change on temperature and evaporation
from a large reservoir in Australia
Fernanda Helfer,*, Charles Lemckert, Hong Zhang
Griffith School of Engineering, Griffith University, Gold Coast, QLD, Australia
* Corresponding author address: Building G09, Room 1.02, Griffith School of Engineering,
Gold Coast Campus, Griffith University, QLD 4222. Tel.: +61 (0)7 5552 7608; Fax: +61 (0)7
5552 8065. E-mail: [email protected]
Abstract: Determining evaporation rates is essential for efficient management of reservoirs
and water resources, particularly in water-scarce countries such as Australia. Today, it is
estimated that open water reservoirs in Australia lose around 40% of their total water storage
capacity per year to evaporation. While this loss is of significant concern, the threat of a
changing climate has been directing greater focus to how much water will be lost from
Australia’s reservoirs in the future. This paper analyses evaporation rates from a large water
supply reservoir in South-East Queensland (SEQ), Australia, under current climate and
predicted climate change conditions using modelling. Daily meteorological projections from
nine global climate models were used in the model DYRESM as the driving forces of the
thermodynamics of the reservoir under study. Two future 20-year period simulations were
undertaken, one from 2030 to 2050, and the other from 2070 to 2090. The modelled future
evaporation rates, as well as water temperatures, were then compared with modelled
evaporation rates and temperatures obtained using observed meteorological variables for
the period of 1990-2010. The results showed that the evaporation rates from the study
reservoir will increase in the future. For the period centred in 2040, the annual evaporation
will be approximately 8% higher than the 20-year average annual evaporation estimated for
the present climate. A more pronounced increase in evaporation is expected in 2070-2090,
with annual evaporation predictions being approximately 15% higher than the baseline
evaporation. The main agent behind this increase is higher surface air temperatures in the
1
future. According to the modelling results, the mean annual surface air temperature will grow
from the present value of 20.4oC to 21.5oC in 2030-2050, and to 23.2oC in 2070-2090. As a
consequence, the mean annual surface water temperatures of the reservoir will increase by
0.9oC and 1.7oC in both timeframes, respectively. This will have a significant impact on the
evaporation rates, particularly in spring and summer, when the temperature increases will be
more significant.
Keywords: climate change, evaporation, reservoirs, temperature
1. Introduction
Australia and most semi-arid countries around the world rely on water stored in reservoirs for
drinking water supply and food production. However, the rates of evaporation in these
countries can be exceedingly large. In Australia, around 40% of its total water storage
capacity is lost per year due to high rates of evaporation (Craig et al., 2005). Concerns over
Australia’s future open water evaporation rates are increasing due to the threat of a
changing climate. In fact, there are indications that the meteorological factors involved in the
process of evaporation will be significantly affected as a result of increasing greenhouse gas
(GHG) emissions. The 2007 Australian Climate Change Report by the Commonwealth
Scientific and Industrial Research Organisation (CSIRO) and Bureau of Meteorology (BoM)
is probably the best source of information on how the Australian climate will change from
now, and into the future (CSIRO and BoM, 2007). This report is based on results from 23
different global climate models used within the Intergovernmental Panel on Climate Change
Fourth Assessment Report (IPCC, 2007). Projections for most climatic variables are given
for a range of GHG emission scenarios. According to the CSIRO and BoM (2007), there is a
growing body of evidence in support of an increase in Australian air temperatures, with an
increase of 0.9oC since 1950. For the future, the estimate of warming by 2030 relative to the
base climate of 1990 is approximately 1.0oC, with warming of around 0.8oC in coastal areas
2
and 1.1oC inland. By 2050, warming in Australia will range from 1.2oC to 2.2oC for low and
high emission scenarios, respectively, and by 2070, from around 1.8oC to 3.4oC.
As a result of the change in climate, particularly the increase in surface air temperatures,
evaporation is also expected to increase throughout Australia. According to CSIRO and BoM
(2007), the largest increases in potential evapotranspiration are expected in the north and
east, where the changes by 2030 are estimated to be around 2%. By 2070, the low emission
scenario shows increases in evapotranspiration of around 3% in the south and west, and
around 6% in the north and east, while the high emission scenario shows increases of
around 6% in the south and west, and 10% in the north and east. Johnson and Sharma
(2010) analysed evaporation changes over Australia using outputs from five global climate
models and the Penman model for open water evaporation (Brutsaert, 1982). The mean
percentage change in open water body evaporation across Australia was found to be 2.2%
for 2030, 3.9% for 2050 and 6.8% for 2070 in a high emission scenario, and 1.7%, 3.9%,
and 4.8% for 2030, 2050 and 2070, respectively, in a low emission scenario
Despite this overall picture of future potential evapotranspiration in Australia, the effects of
climate change on evaporation from and thermodynamics of particular lakes can only be
determined by performing a site-specific investigation. This can be undertaken using sitespecific projections for the meteorological variables that drive the evaporation process,
namely solar radiation, air temperature, wind speed and vapour pressure. Evaporation under
climate change in Australia has only been assessed on a broad scale, based on
meteorological projections generated by global climate models. These do not incorporate
local climate variations, and therefore are inappropriate for finer scale analyses, such as the
study of evaporation from reservoirs. Local studies on evaporation should be based on
meteorological data derived specifically for the study site and should incorporate the regional
pattern variations of the local climate.
3
After obtaining the appropriate meteorological data, the study of the response of lakes to
climate change can be achieved by using numerical modelling simulation. In the modelling of
lakes under climate change conditions, the predicted future thermodynamics and
evaporation driving forces are incorporated in the model, and the responses of the lakes to
these forces are obtained in a relatively short period of time. This further facilitates the
investigation of the behaviour of lakes under different climate change scenarios and allows a
portfolio of all possible outcomes to be created.
Daily climate change projections into the current century for any country can be obtained
from global climate models, as mentioned earlier. Global climate models, also known as
general circulation models (GCMs), are models that solve the primitive equations of mass,
momentum and thermodynamics to generate a description of the state of the atmosphere,
and produce most of the meteorological variables, such as wind speed, relative humidity,
rainfall, surface air temperature and solar radiation. The GCMs are typically used together,
as a multi-model analysis, where long time series into the future and different GHG emission
scenarios are used in order to create climatic statistics (Nunez and McGregor, 2007). This
multi-model approach allows for a first look at the likability of future climate changes, which
is usually referred to as climate change ensemble projection or climate change “envelope”.
Inherent to GCMs, however, is their coarse spatial resolution which, amongst all the
available models, varies from 200 to 500 kilometres. With this resolution, albeit GCMs
incorporate the important large-scale atmospheric circulation, they are unable to capture
local-scale factors, such as the orographic elevation, proximity to water bodies and local
winds. One way to work around this is to downscale the GCM datasets to a regional scale
using a regional climate model (RCM) with atmospheric forcings based on the outputs of the
GCM. In this way, the local climate trends are captured, while maintaining agreement with
the large-scale response of each GCM (Thatcher and McGregor, 2011). For instance, Sahoo
and Schladow (2008) used meteorological projections from the global climate model GFDL
4
(Delworth et al., 2006) downscaled to the California-Nevada region to study the impacts of
climate change on the dynamics of lakes and reservoirs in that region. Likewise, McGregor
and Nguyen (2010) used downscaled projections forced by sea-surface temperatures from
the CSIRO-Mk 3.5 model (Gordon et al., 2010) to study the future climate of the MurrayDarling Basin, Australia. Furthermore, Nguyen et al. (2011) utilised the downscaling
technique for six global climate models to investigate the local climate of the Pacific island
nations. Also, Yilmaz and Imteaz (2011) downscaled data from the GCM ECHAM5 in order
to simulate snowmelt runoff from upper Euphrates basin. According to these authors, highresolution simulations provide a better resolution of spatial details and of extremes, in
addition to yielding boundary conditions for even finer resolution simulations.
Kay and Davies (2008), for example, investigated the differences in the estimation of
monthly potential evaporation in Britain when using outputs from either GCMs or RCMs.
They utilised the Penman-Monteith model and a simple temperature-based model in the
estimations of evaporation for the current climate. The results were compared to a gridded
dataset derived from a modified Penman-Monteith formulation using meteorological
observations. According to their results, RCMs’ outputs are able to generate monthly
evaporation rates that show much closer agreement with evaporation rates derived from
observed data as compared to GCMs’ outputs.
In Australia, the CSIRO has converted broad-scale climate change projections from a wide
range of GCMs into local-scale regional projections for the South-East Queensland (SEQ)
region using the regional climate model CSIRO Conformal-Cubic Atmospheric Model
(CCAM) (McGregor, 2005; McGregor and Dix, 2001; McGregor and Dix, 2008). The finer
scale detail of these downscaled projections varies from 15 km to 60 km and thus they are
more reflective of regional considerations than the popular GCM outputs, and thereby more
representative of a plausible future.
5
In this context, the aim of this paper was to investigate the influences of a changing climate
on evaporation from a large and important water body located in SEQ, Australia.
Meteorological projections from nine GCMs, subsequent to the CCAM dynamical
downscaling, were used in a validated dynamic reservoir model (Helfer et al., 2011).
Modelled future volume-averaged lake temperatures, lake surface temperatures and
evaporation rates were analysed and compared with the present-day modelled temperatures
and evaporation rates. Also, causes and implications of lake temperature and evaporation
changes are briefly discussed. The outcomes of this research provide a first indication of the
most likely changes in evaporation and water temperatures expected for reservoirs in the
SEQ region for all seasons, as well as a better understanding of the drivers of these
changes. This study also provides a warning of the possible changes for which the region
should be prepared in terms of water availability in the future.
2. Study site and meteorological data for simulations
For this study, a large reservoir located in SEQ was considered. Wivenhoe Dam
(27°23'11"S, 152°37'10"E, Figure 1) is built on the Brisbane River and is located
approximately 40 km northwest of Brisbane. Its main purposes are flood mitigation and the
supply of potable water to the SEQ region. At full capacity, this dam has a volume of 1,160
hm3 and a surface area of 107 km2, with a maximum depth of 40 metres.
Figure 1.
The meteorological data for the nearest official weather station, located 25 km from
Wivenhoe Dam, was provided by the BoM. This was the nearest station with a considerably
long, reliable and consistent set of historical meteorological data. Daily data from 01/01/1990
to 31/12/2010 were used in the simulations to compose the baseline (or present-day)
scenario. This long-term set of data allowed for an accurate estimation of the present-day
average thermal behaviour and evaporation of the reservoir.
6
Future daily projections for solar radiation, wind speed, air temperature, air humidity and
rainfall over the SEQ region were provided by the CSIRO. Although there are currently 23
GCMs available from the World Climate Research Programme (WCRP) Coupled Model
Intercomparison Project phase 3 (CMIP3) multi-model dataset archive (Meehl et al., 2007),
only nine GCMs were considered by CSIRO for downscaling in SEQ (Table 1). The selection
of these GCMs was predicated on the models’ efficacy in predicting the present-day climate,
particularly the seasonal variability, throughout Australia. The nine host GCMs were used to
derive daily data on a 15-km resolution grid over the SEQ region using dynamical
downscaling with the Conformal-Cubic Atmospheric Model (CCAM). The downscaling
technique is described in Nguyen et al. (2011). Meteorological data at the location
27°22'12"S, 152°36'00"E was extracted from the grid. This point coincides with the centre of
the southern portion of the dam, where the greatest depth occurs.
The efficacy of climate models in simulating the current climate has been represented in
literature by different skill scores. Perhaps one of the most reasonable ranking systems
available for daily projections is the one proposed by Perkins et al. (2007). These authors
ranked all the available GCMs based on how well each model represents the observed
probability density functions of daily precipitation and maximum and minimum surface air
temperatures in Australia. Watterson (1996), on the other hand, ranked a larger range of
GCMs based on the agreement between modelled and observed seasonal-mean
temperature, precipitation and sea level pressure in Australia. In this system, a single M
value between 1 (perfect match) and 0 (no-skill) is obtained for each model, variable and
season, based on pattern correlations and root-mean-square errors. It is on the M score,
rather than on the score by Perkins et al. (2007) that the choice of the GCMs utilised in this
study was based. This is due to a larger range of models covered by this system. The M
value of each model is presented in Table 1. The value represents the mean skill score for
each of the three aforementioned meteorological variables for each of the four seasons (that
7
is, a mean of a total of 12 scores). The mean skill scores of the selected models are all
above 0.6, meaning that they have high ability to predict the Australian climate. In this study,
a method of weighting the models’ performance based on the M scores, as a means to
obtain a clearer sign of the likely change (Kirono and Kent, 2011) was also considered, as
described in section 5.
Table 1.
2.1. Bias correction
In this section, the approach to remove the biases (errors) from the projections derived from
the nine models used in this study is described. The bias correction method is based on the
delta change approach, a method that has been extensively discussed and applied in similar
studies based on climate model predictions (eg, Hay et al., 2000; Lenderink et al., 2007; van
Roosmalen et al., 2010; Xu & Yang, 2012).
As described earlier, the time period from 01/01/1990 to 31/12/2010 was adopted as the
baseline (or present-day) scenario in this study. Observed daily meteorological variables
(precipitation, air temperature, solar radiation, wind speed and relative humidity) for this
period were first compared against the meteorological variables derived from the nine
climate models for the same period. Even though all models reproduced the seasonal cycle
of the five meteorological variables quite well (particularly surface air temperature, solar
radiation and humidity), almost all of the models showed significant discrepancies at some
time for some particular months and for some variables, in comparison to the observed data.
For example, in general, all models underestimated solar radiation and air temperature in all
months. Also, the models significantly overestimated precipitation, particularly between
March and August. They also overestimated wind speed in spring. As for vapour pressure,
there was a generalised underestimation in summer. These biases were removed using the
statistical approach described below.
8
The delta change bias removal approach is a statistical correction method which aims to
produce long-term time-series that have a monthly (or seasonal) statistical mean close to
that of the observations. The standard delta change method consists of computing long-term
averages of monthly (seasonal) differences between the modelled time-series in the control
scenario and the modelled time-series in the future scenario. The method assumes that
these biases (ie, the monthly or seasonal differences between present and future modelled
data) do not change over time. In order to produce a future time-series of the meteorological
variables, these calculated biases are applied to the observed historical climate data. The
delta method thus assumes that future model biases for both mean and variability will be the
same as those in present-day simulations.
In this study, an adapted approach based on the delta change method was used. In this
approach, the biases were calculated for the present-day scenario as being the difference
between the modelled and the observed data. These changes were then subtracted from the
modelled future daily time-series. More precisely, instead of creating databases for the future
by altering an observed current database of meteorological variables with delta change
factors, the modelled future daily databases were altered with the delta change factors
obtained from the mean difference between present-day modelled and observed time-series.
This is justifiable because future data on a daily basis was available for this study. The
standard method would be more appropriate in the case where future data is only available
on a monthly basis, and there is a need to obtain daily data series for future simulations.
Another innovation of the approach utilised in this study is the use of the running monthly
mean value instead of the simple monthly mean in the estimation of the delta change
factors. This method was suggested by Haerter et al. (2011) to avoid discontinuities at the
interfaces between months. For state variables, such as relative humidity and surface air
temperature, the delta changes (Δs) were computed for each day i as:
9
s(i)  Vm, p (i)  Vo, p (i) ; i = 1, 2, ... , 365
(1)
where the indices ‘m’ and ‘o’ stand for modelled and observed data respectively, and the
index ‘p’ refers to the reference period (1991-2010). V (i ) is the running 31-day mean value
for all 20 years of the reference period, and involves the previous and subsequent 15 days to
avoid discontinuities at the interfaces between calendar months, and is given as:
 1 i 15
; 15  i  350

 V ( j)
 31 j i 15

365

 1  i 15
V (i )     V ( j )   V ( j )  ; 1  i  15
j 365  (15 i )

 31  j 1
  365
15 (365 i )
 1   V ( j )   V ( j )  ; 350 < i  365
 31  j i 15
j 1

(2)
For flux variables, such as solar radiation, wind speed and precipitation, the delta changes
(Δf) were computed for each day as:
f (i ) 
Vo, p (i )
Vm, p (i )
; i = 1, 2, ... , 365
(3)
For leap years, an additional delta change factor was used after day 59. The value of this
factor was considered as the same as the one calculated for day 59.
To remove the biases from the future modelled time-series, the following procedures were
used. For relative humidity and air temperature, the new databases were generated as:
Vc, f (i)  Vm, f (i)  s(i) ; i = 1, 2, ... , 365
(4)
10
where Vc , f (i ) represents each daily value in the future corrected databases and Vm, f (i ) is
the original modelled daily value in the uncorrected databases. For solar radiation, wind
speed and precipitation, the procedure was:
Vc , f (i )  Vm, f (i ) f (i ) ; i = 1, 2, ... , 365
(5)
The method of bias correction was tested with the observed and modelled daily data series
for the reference time period (1991-2010). The method significantly improved the data
series, as shown in Figure 2. For instance, the mean biases in monthly solar radiation were
reduced from -3.3% to -0.09%. Vapour pressure mean biases decreased from -4.3% to 0.09%. For monthly wind speed, the errors were reduced from 3.8% to -0.04%. The mean
biases in air temperature were reduced from -6% to -0.02%, and in rainfall, from 40% to
0.3%.
3. Simulations
All simulations to estimate Wivenhoe Dam’s temperatures and evaporation rates were
performed with the 1-D model DYRESM. This hydrodynamic model simulates lake
thermodynamic processes and lake surface mass fluxes using meteorological variables,
inflows and outflows as driving forces. The model is described in brief in the following
section.
For the present-day simulation, daily observed meteorological variables from 1990 to 2010
were used in the model. For the future simulations, two distinct 20-year periods were
chosen, one around 2040 and one around 2080. The daily future meteorological projections
were obtained from 9 climate models and passed through a bias correction algorithm to
reduce systematic errors, as described in the previous section.
11
The future scenarios were simulated in the SRES-A2 emission scenario (IPCC, 2000). The
SRES-A2 scenario comprises a high GHG emission scenario in which the concentration of
CO2 reaches 850 ppmv (parts per million by volume) by 2100. As current emission levels are
at, or are already above those specified for this scenario (Raupach et al., 2007), the SRESA2 scenario appeared to be the most realistic choice. Other similar studies, such as Kay et
al. (2006), Charles et al. (2007), Kay and Davies (2008), Leander et al. (2008), Chu et al.
(2010), Nguyen et al. (2011) and Yilmaz and Imteaz (2011) have also opted for this
scenario.
Figure 2.
The initial depth of Wivenhoe Dam was set at full capacity (40 metres) and water column
temperatures were assumed to be the same temperatures as measured in January 2007.
The very high similarity between the air temperatures in 2007 and 1990 was the main reason
for this approach, which implies that the lake had similar water temperatures in January for
those two years. As these values only approximate actual initial conditions, a model warmup period was considered. Previous sensitivity tests have indicated that Wivenhoe Dam
reaches equilibrium with external forcings within one month (Helfer et al., 2011).
Also for the simulations, vertical mixing processes were assumed to be more important than
horizontal advective processes, such as inflows and outflows, in determining the vertical
temperature distribution in the study reservoir. In large reservoirs, such as Wivenhoe Dam,
where inflows and outflows have only localized effects on water temperature, these flows
may be ignored when predicting the temperature in the central portion of the lake (Ivey and
Patterson, 1984). Other studies on the temperature dynamics of Wivenhoe Dam (eg, Yao,
2008; Helfer et al., 2011) have found close agreement between modelled and measured
water temperatures without considering inflows and outflows. Therefore, the only water
contribution to the dam was the rainfall, and the outflow was the evaporation.
12
4. Modelling with DYRESM
In this study, the one-dimensional processed-based model DYRESM was used to model the
lake mixing dynamics and to predict water temperatures and evaporation rates from
Wivenhoe Dam. The model was previously calibrated for Wivenhoe Dam against measured
water temperatures, and the validation for evaporation was based on measured data
obtained for a nearby dam. Full details of the validation and calibration of the model are
described in Helfer et al. (2011).
An extensive description of DYRESM has been given in literature (eg, Imberger et al., 1978;
Spigel and Imberger, 1980; Imberger and Patterson, 1981; Patterson et al., 1984; Hocking et
al., 1988; Patterson and Imberger, 1989). In brief, DYRESM takes daily or sub-daily
meteorological forcing data (solar radiation, wind, surface air temperature and air humidity),
volume of inflows and outflows, and then produces outputs for water depth, water
temperature, salinity and density. The model assumes horizontal homogeneity, which is
based on the lake density stratification. Vertical motions are inhibited by this stratification,
while horizontal variations in density are quickly relaxed by horizontal advection and
convection.
The basic processes modelled by DYRESM are surface heat and mass transfer, dynamics
of the surface layer, vertical diffusion in the hypolimnion, and inflow and outflow dynamics.
The model makes use of the Lagrangian layer concept in which the lake is represented by a
series of horizontal layers, each with uniform properties but variable thickness. These layers
are able to expand, contract, amalgamate, subdivide and move up and down as they are
affected by the inflows, outflows, evaporation and rainfall.
Within the scope of this paper, the processes which are most important are those controlling
the surface layer. The surface heat, mass and momentum exchanges comprise the primary
13
mechanisms for heating, mixing and stratifying a water body in DYRESM. The surface layer
dynamics is based on an integral turbulent kinetic energy model (Sherman et al., 1978;
Yeates and Imberger, 2003). The turbulent kinetic energy budget is computed through three
main processes: convective overturn (where energy is released from the decrease in
potential energy resulting from dense water falling to a lower level), stirring (where energy
from the wind stress is applied to the surface layer), and shear (where kinetic energy is
transferred from upper to the lower layers in the water column). The total available energy
and the required potential energy for mixing the surface mixed layer and the adjacent layer
are compared. If the energy available is greater than the energy required for mixing, the
mixed layer is deepened and the available energy is decremented by the required energy.
The process is repeated until there is no remaining energy to continue the deepening
process.
Whenever layers are mixed together, the layer properties are redistributed according to the
conservation laws:
Cnew 
Ci M i  Ci 1 M i 1
M i  M i 1
(6)
where the subscripts refer to layer indices, C is the property being conserved (energy, salt or
momentum) and M is the layer mass.
The surface heat fluxes include heating due to shortwave and long-wave radiation, sensible
heat and latent heat. The fluxes of long-wave radiation, sensible heat and evaporation are
assumed to operate on the surface layer only. Short-wave radiation is assumed to act not
only on the surface layer, but also in the water column, penetrating the water depth
according to the Beer-Lambert law, as a function of the light extinction coefficient. The
quantity of radiation reaching the surface of the lake depends on the albedo of the water.
The penetrative radiation is assumed to be 45% of the radiation reaching the surface. Long
14
wave radiation emitted from the lake is calculated according to the Stefan-Boltzmann law, as
a function of the emissivity and temperature of the water surface.
The sensible and latent heat fluxes are described by bulk aerodynamic formulae (Brutsaert,
1982). For the evaporative heat flux, which is most important for the present discussion, the
bulk aerodynamic formula is given as:
QE  a CEU qa  qs 
(7)
where QE is the latent heat flux due to evaporation (W m-2), ρa is the air density, λ is the
latent heat of evaporation, CE is the latent heat transfer coefficient, U is the wind speed at
the reference height of 10 m, qa is the specific humidity in the air and qs is the specific
humidity at saturation pressure, which is a function of the surface water temperature.
The surface mass exchanges include mass input from precipitation and output from
evaporation. The mass of the evaporated water is calculated as:
MN  
QE AN

(8)
where MN is the evaporating mass from the surface layer with area AN. The total change in
water mass in the surface layer in each time interval is given by the difference between the
rainfall and the evaporation.
15
5. Results and discussion
5.1. Present-day scenario (1991-2010)
This section describes the average climate of the baseline scenario, giving emphasis to the
driving forces of evaporation (surface air temperature, solar radiation, humidity and wind
speed). The results of modelled lake temperature and evaporation in the present-day
scenario are then discussed. Being the model warm-up period, the year of 1990 is not
included in these analyses.
Figure 3 presents the seasonal variability in surface air temperature, wind speed, solar
radiation and vapour pressure at Wivenhoe Dam for the 20-year set of data from 1991 to
2010. These variables were selected for presentation due to their relevance in the process of
evaporation. The whiskers show the year-to-year variability and range from the 10th to the
90th percentiles of the observations.
Figure 3.
As presented in Figure 3, the Wivenhoe Dam region is characterized by four well-defined
seasons, with hot, humid summers and relatively cold, dry winters. The mean annual
temperature is around 20oC, with the air temperature in summer being around 25oC and in
winter, 15oC. These temperatures characterize a typical temperate climate. Solar radiation at
Wivenhoe Dam is also seasonal, with higher values in summer than winter. Similarly, vapour
pressure follows the same trend as air temperature and solar radiation, with high values in
summer and low values in winter. The deficit of vapour pressure (not shown in Figure 3) is
also higher in summer, indicating that higher rates of evaporation are expected during this
season. Wind speed, on the other hand, does not follow a seasonal trend. The average wind
speed over the 20-year period was 3 m s-1.
16
The modelled 20-year average seasonal daily evaporation from Wivenhoe Dam is presented
in Figure 4. There is a clear difference between evaporation in winter and summer. The 20year average daily evaporation is 3.6 mm. Daily evaporation in summer is around 5 mm,
whereas in winter it is around 2 mm. In spring and summer, the daily evaporation is about
3.6 mm. In summer, the 20-year average total evaporation is 450 mm and in winter it is 190
mm. Evaporation in spring and autumn falls between these two extremes, at around 330
mm. The 20-year average evaporation from the dam is 1,300 mm, varying from 1,080 to
1,490 mm.
Figure 4.
The modelled temperature of Wivenhoe Dam for the period of simulation is presented in
Figure 5. The variation in water depth over the years has been omitted in the Figure, as this
is a free-surface representation of the lake’s temperatures. There is a notably consistent
pattern in the variation of the lake temperatures, with high temperatures in summer and low
temperatures in winter. This pattern is a result of the seasonal variation in the surface air
temperatures and solar radiation shown previously in Figure 3. The 20-year average lake
surface temperature is 22.4oC, with an average of 28.2oC in summer and 15.6oC in winter.
The lake bottom temperature is nearly constant in Wivenhoe Dam, with an average of
15.2oC, oscillating between 15.4oC in summer and 14.8oC in winter. The 20-year average
volume-averaged lake temperature is 19.1oC, varying from an average of 22.1oC in summer
and 15.3oC in winter.
Figure 5.
Also notable in Figure 5 is the pattern of thermal stratification in Wivenhoe Dam. Wivenhoe
Dam is a warm monomictic lake that stratifies from September until the beginning of May,
17
covering spring, summer and the autumn seasons. During winter, Wivenhoe Dam is
characterized by well-mixed water columns due to the lake’s turnover. Figure 6 shows
measured in-lake temperature profiles during the year of 2007. Field measurements were
obtained from SEQ Water.
Figure 6.
Thermal stratification is considered an important regulator of the overall metabolism of a
lake. The epilimnion – upper and warm water layer – is usually well mixed as it is subject to
mixing induced by the wind. On the other hand, the hypolimnion – bottom layer of colder
water – experiences limited mixing because it is isolated from surface energy inputs due to
the existence of a thermocline region between itself and the epilimnion. This region, called
metalimnion, limits the exchange of dissolved substances between the epilimnion and the
hypolimnion due to the lowering of turbulence and mixing. Limited mixing has important
implications for the cycling of critical constituents such as nutrients and dissolved oxygen. As
shown in Figure 6, in Wivenhoe Dam, the metalimnion is present during almost the entire
yearly cycle, with the highest thermal gradients occurring from November to April. The depth
of the epilimnion (or depth of the well mixed layer) is indicative of the level of mixing in the
lake. In Wivenhoe Dam, the depth of the mixed layer has an average of 3.6 metres in
summer, 6.8 metres in autumn, 5.0 metres in spring and it is well mixed from top to bottom
during winter.
5.2. Future scenarios
This section firstly describes the future climate at Wivenhoe Dam, pointing out the main
changes in the driving forces of evaporation – namely air temperature, wind speed, solar
radiation and air vapour pressure. Assessment of the causes of these climate changes,
however, is not within the scope of this paper. The results of future modelled evaporation are
then discussed, followed by a discussion on changes in lake temperature.
18
The modelled results under climate change are presented in an ensemble projection of the
nine models used in this study, allowing for an assessment of the uncertainties implicit in the
choice of the GCMs. The results are represented by the average of the outcomes from the
nine climate models and the variability is indicated by the range between the 10th and the
90th percentile values, following similar studies (e.g. CSIRO and BoM, 2007 and Smith and
Chandler, 2010).
Driving forces of evaporation under climate change
The 20-year average seasonal air temperature, wind speed, solar radiation and air vapour
pressure projected for the periods 2030-2050 and 2070-2090 are summarized in Figure 7.
The red and green series represent the mean of the 20-year averages derived from the nine
models, and the whiskers illustrate the intervals between the 10th and the 90th percentiles of
the outcomes. The blue circles show the 20-year average values observed in the presentday scenario (1991-2010).
Figure 7.
The data presented in Figure 7 shows a slight increase of average annual surface air
temperature in the period of 2030-2050 in relation to the present-day average climate. This
increase is predicted to be approximately 1oC, ranging between 0.8oC and 1.3oC, depending
on the climate model. The most prominent rise in temperature for this period occurs in
spring, when air temperatures are likely to increase by around 2oC, representing a rise of
10% in relation to the present climate temperatures. The least significant rise is in autumn,
when the average air temperature is expected to rise by an average of 0.7%. For 2070-2090,
again all the models agree with a growth of the average air temperature in all seasons at the
Wivenhoe Dam area. However, the estimated air temperature increases for this timeframe
are more significant than the increases estimated for 2030-2050, lying between 1.7 and
19
3.2oC (average of models = 2.7oC). Higher air temperature increases are expected in spring
and winter, with a rise ranging between 3 and 4oC depending on the model. This represents
an increase of approximately 20% in comparison with the baseline scenario air
temperatures.
The average annual solar radiation is not expected to change significantly in 2030-2050 and
2070-2090 in relation to the present-day scenario. There is a slight rise in radiation for both
timeframes, of around 0.5%. However, the estimated changes vary from -0.8% to 1.5%
between the climate models, which makes it difficult to draw a definite conclusion about
future annual solar radiation. However, seasonal changes in solar radiation are evident in
both timeframes, as shown in the bar graphs in Figure 7. Solar radiation is likely to go down
in summer and autumn, but go up in winter and spring, in comparison to the baseline solar
radiation. For 2030-2050, the range of change between models’ results for spring and winter
solar radiation is between 3% and 7% (average of models = 5%). For 2070-2080, this range
is between 6% and 12%, with an average of 8% for both seasons. The decline in solar
radiation in summer and autumn for the 2030-2050 timeframe is likely to be -2% and -5%,
respectively. When looking at the 2070-2090 timeframe, the decrease in summer solar
radiation is expected to be -4% and in autumn solar radiation, -11%.
According to the nine climate change models, future air vapour pressure is anticipated to be
just slightly higher than the current long-term vapour pressure in all seasons and in both
timeframes. The average increase in vapour pressure is approximately 0.5% in 2030-2050
and 1.2% in 2070-2090, with a range of change between models varying from -0.4% to 2%
in 2030-2050 and from -0.9% to 4% in 2070-2090. An increase in average vapour pressure
is expected in summer, winter and spring, for both periods of time. In summer, this growth is
expected to be of around 1% and 1.4%, respectively for 2030-2050 and 2070-2090. In
spring, it is expected to be slightly higher than in summer, at 1.2% and 3.2% for both
timeframes respectively. In winter, the increase is in the order of only 0.3% for both time
20
periods. As opposed to the other seasons, in autumn, vapour pressure will experience a
small decrease, in the order of 0.3% in both timeframes.
The long-term average wind speed in the future projections is estimated to be just slightly
lower than the long-term average wind speed observed in the present-day scenario. For
2030-2050, the average annual wind speed is projected to drop 1.6%. For 2070-2090, it is
likely to be 1.3% lower than today’s wind speed. As for seasonality, no pronounced changes
in wind speed are estimated in future summers. Wind magnitude is expected to slightly
increase in spring, with the average growth from the nine models being 0.5% and 4.4% for
the timeframes of 2030-2050 and 2070-2090, respectively. As opposed to spring and
summer, considerable reductions in wind speed are expected in winter, with drops in the
order of -5.2% and -8.6% for 2030-2050 and 2070-2090, respectively. Wind speed is also
estimated to drop in future autumns. In the time period of 2030-2050, the expected decrease
is approximately -1.5%, and in 2070-2090, -3.4%.
Figure 8 shows the inter-model variability for the annual projections of air temperature, solar
radiation, vapour pressure and wind speed. The red solid line represents the current 20-year
annual average values. The grey bars show the 20-year average of the modelling data and
the whiskers range between the 10th and the 90th percentiles. It can be seen that all the
models predict a rise in air temperatures and that this rise is higher in the 2070-2090
timeframe. As for solar radiation, eight models predict an increase in solar radiation, but this
increase is just slight for both timeframes. Similarly, average annual vapour pressure is also
predicted by eight models to increase slightly. This increase is more significant in the 20702090 timeframe. The average annual wind speed is predicted by six models to be slightly
less than the current 20-year average.
Figure 8.
21
Evaporation under climate change
As a result of the aforementioned changes in climate, evaporation from Wivenhoe Dam is
also expected to change, as is illustrated in Figure 9.
Figure 9.
Compared to the baseline evaporation, the average annual evaporation in the period 20302050 is likely to be 5.6% higher, and in the period of 2070-2090, 14.5% higher. In 20302050, annual evaporation is estimated to be 1,400 mm, compared to the long-term annual
evaporation in the present-day scenario of 1,300 mm. It is important to note that all nine
simulations predict higher evaporation rates than the current long-term average, as shown in
Figure 10. The variation in long-term annual evaporation for the 2030-2050 period ranges
from 1,330 mm to 1,410 mm between the models. When looking at the 2070-2090 period of
simulation, the average annual evaporation is 1,490 mm, with variation between the models
ranging from 1,410 mm to 1,540 mm. These increases are in accordance with the findings of
Johnson and Sharma (2010), which also showed increase in average evaporation in
Australia in 2030, 2050 and 2070 under the SRES-A2 scenario. They used outputs of five
GCMs and the Penman equation to predict future evaporation rates over Australia.
Figure 10.
Evaporation is expected to increase in summer, winter and spring in the future scenarios,
with this rise being more pronounced in 2070-2090 than in 2030-2050. The more significant
increases in evaporation are expected in spring, with an average increase of 21% in 20302050 and of 40% in 2070-2090. It is important to note that all considered models, with no
exception, predict a significant rise in spring evaporation. The models’ predictions vary
between 13% and 26% for the 2030-2050 timeframe, and between 30% and 48% for 20702090. The main reasons behind the increase in spring evaporation – based on the expected
22
changes in the climatic variables presented in Figure 7 - are the significant increases in
surface air temperatures, which are likely to rise by around 2oC in 2030-2050 and by 4oC in
2070-2090, solar radiation and the slight increases in wind speed.
In winter, evaporation is expected to increase by 1% in 2030-2050 and by 14% in 20702080. In the 2030-2050 timeframe, the results using the nine climate models showed an
inter-variation ranging from -2% to 4%. Therefore, some of the models predict a decrease in
winter evaporation, which can be attributed to the significant reductions in wind speed.
These reductions may outweigh the effects of increasing air temperatures and solar radiation
during this season. Nevertheless, most of the models predict a rise in winter evaporation,
resulting in an average increase of 1% in relation to the baseline scenario. In the 2070-2090
timeframe, on the other hand, all models predict higher evaporation rates in winter, varying
from 9% to 21% above the baseline values. The main cause of these increases is the
significant elevation in air temperatures in winter. Even with lower wind speeds, winter
evaporation in 2070-2090 is still greater than winter evaporation in the control scenario due
to the higher air temperatures.
In summer, evaporation is expected to increase by 5% in 2030-2050 and by 12% in 20702090. Increased summer evaporation is estimated by all of the models, and the main cause
of this rise is the elevation in air temperatures. Even with predicted decreases in wind speed
and solar radiation, and slight increases in vapour pressure, the resulting effect is still an
increase in evaporation in summer, attributable to the higher air temperatures.
Autumn is the only season when evaporation is predicted to be lower than the current longterm evaporation. The reduction is predicted to be, on average, -5.5% in 2030-2050 and -8%
in 2070-2090. All the models adopted in this study resulted in lowering of evaporation rates
in autumn. The variation between the models is between -2.7% and -9% in the 2030-2050
timeframe, and between -3.5% and -12% in the 2070-2090 timeframe. The main factors
23
leading to reduced evaporation rates in this season are the reductions in solar radiation and
wind speed. The effects from these reductions on evaporation outweighed the effects of the
slight increases in air temperatures, as shown in Figure 7.
Figure 11 shows the predictions after weighting the evaporation results from the nine models
based on the M score for each model. This score, as mentioned earlier, is based on the
agreement between simulated and observed seasonal-mean air temperatures, precipitation
and sea level pressure across Australia. Even though this score is not based on evaporation
projections, it is still one of the closest available forms of assessing the quality of the
projections. Models with higher M value were assumed to give more reliable projections of
future climate, and therefore were assigned a higher weight (eg, 27% for ECHAM5, and 21%
for CCSM3, which predict the current average climate with higher level of accuracy in
comparison with the other models).
Figure 11.
As noted in Figures 11 and 9, the results with the model weighting system did not differ from
the results obtained without a weighting system. According to both methods, higher
evaporation rates are expected in future summers and springs and lower evaporation in
future autumns. In winter, evaporation is only significantly higher than the baseline values in
the 2070-2090 timeframe. Both methods also yielded similar magnitudes for the future
annual evaporation rates.
Lake temperature under climate change
The changes in climate presented in Figure 7 also affect Wivenhoe Dam temperatures, as
shown in Figure 12. Overall, it can be seen that there is an increase in the long-term average
surface temperature of Wivenhoe Dam in the future scenarios in comparison to the baseline
long-term average lake surface temperature. The average lake surface temperature
24
increases from 22.4oC (baseline value) to 23.1oC in 2030-2050, and to 24.05oC in 20702090. All of the models predict an increase in the average lake surface temperature (Figure
13). In 2030-2050, the variation between the outcomes from the nine models lies between
0.55oC and 0.80oC, whereas in 2070-2090, it lies between 1oC and 2oC.
Figure 12.
Figure 13.
Lake surface temperatures are expected to increase more significantly in spring. For the
2030-2050 timeframe, the increase is expected to be 1.7oC in relation to the baseline spring
lake surface temperatures. For the 2070-2090 timeframe, the increase is even higher, at
3.3oC in comparison to today’s spring lake temperatures. Also significant are the changes
expected in winter. In this season, lake surface temperatures in 2030-2050 are estimated to
be 1oC higher than the baseline winter surface temperatures, and 2.5oC higher in 20702090. In summer there is also a predicted increase in lake surface temperatures, but this is
not as pronounced as the increases estimated for winter and spring. The increase in
summer lake surface temperatures is expected to be 0.4oC in 2030-2050 and 1oC in 20702090 in comparison to the baseline lake temperatures in summer.
The volume-averaged temperature of Wivenhoe Dam is estimated to be approximately 19oC
in the present-day scenario. The simulations with the future climate projections resulted in an
average lake temperature of 19.5oC in 2030-2050 and 20.5oC in 2070-2090. As with lake
surface temperature, the average lake temperature is also expected to increase more
significantly in spring than in the other seasons. The higher water temperatures in spring
explain the more pronounced increase in evaporation in this season.
25
Table 2 summarizes the changes predicted in the meteorological variables used in the
modelling of evaporation, as well as the changes in lake temperature and in evaporation
from Wivenhoe Dam for the two future timeframes adopted in this study (2030-2050 and
2070-2090). The changes are categorised as significant, slight and unchanged. The table
shows that the main agents behind the significant increases in evaporation and lake
temperature in the future timeframes (eg, in spring) are primarily the increased surface air
temperatures. Ii is also noticeable that wind speed plays an important role in the
determination of evaporation rates. Even small increases/decreases in this climatic variable
will cause a significant change in evaporation. For instance, in 2030-2050, the models
predict a significant increase in air temperature and a slight increase in solar radiation in
both winter and spring seasons. In winter, however, there is a slight reduction in wind speed,
whereas in spring, there is no change in wind speed. The slight reduction in wind speed in
winter counterbalanced the increases in air temperature and solar radiation, resulting in
unchanged evaporation rates in this season. In spring, however, as there was no change in
wind speed and therefore, nothing to counterbalance the high air temperatures and solar
radiation, the result was a significant increase in evaporation. This important effect of the
wind can be explained by the aerodynamic model used to estimate evaporation in the model
DYRESM. This model is largely sensitive to changes in this parameter.
Table 2.
Figure 14 shows the average volume-averaged lake water temperature from 1991 to 2010
and for the two future projections. There is a noticeably gradual increase of warming of
Wivenhoe Dam over the years. From 1991 to 2010, the rate of warming was 0.02 oC per
year. From 2030 to 2050, the warming rate is estimated to be approximately 0.03oC per
year. From 2070 to 2090, the rate of warming is likely to reach 0.035oC per year. The impact
of such changes will not only affect evaporation rates, but also the biological activity of the
lake. For instance, higher water temperatures will decrease the maximum amount of oxygen
26
that can be dissolved in the water, leading to oxygen stress if the water receives high loads
of organic matter. This will have a negative impact on the biological processes of fish
including growth, reproduction and behaviour.
Figure 14.
6. Conclusions
This study analysed the changes in water temperatures and evaporation from a large
Australian reservoir under climate change conditions using modelling. The conclusions
drawn here are based on Wivenhoe Dam, a temperate lake located in the state of
Queensland.
The climate change scenario used in this study was the SRES-A2 emission scenario, which
is a realistic scenario based on current global greenhouse gas emissions. Two future
timeframes were used, one from 2030 to 2050 and the other from 2070 to 2090. Projected
daily meteorological data from nine global climate models downscaled to a 15-km resolution
grid over the study area were used to derive daily water column temperatures and
evaporation rates through modelling with the well-established 1-D model DYRESM. The
projected modelled water temperatures and evaporation rates were compared to modelled
water temperatures and evaporation rates obtained with observed meteorological data from
the period of 1991 to 2010. An ensemble of projections was used in this assessment to
account for the uncertainties involved in the choice of climate models.
The simulations showed that the long-term average annual evaporation from Wivenhoe Dam
is 1,300 mm. While the average annual evaporation in 2030-2050 is slightly higher than the
baseline evaporation, at 1,400 mm, and the predicted average annual evaporation in 20702090 is considerably higher, at 1,490 mm. This is in agreement with the work of Johnson
27
and Sharma (2010), who has also demonstrated that evaporation will increase in Australia in
the future.
By observing the future climate against the evaporation predictions, it can be concluded that
the main changes in the driving forces that bring about the increase in evaporation are the
increased surface air temperatures. Higher air temperatures are expected in the future in all
four seasons, particularly in summer, winter and spring, while higher wind speeds are
expected in summer, autumn and spring.
The season which will experience higher changes in evaporation is spring, due to not only
higher air temperatures, but also higher solar radiation and wind speeds. As opposed to
spring, autumn seasons will experience a slight decrease in evaporation, due to decrease in
solar radiation and wind speeds.
Therefore, it appears that evaporation from SEQ reservoirs will increase in the future, but
this increase will not be significantly higher than the baseline evaporation over the next 50
years. Beyond this timeframe however, important increases in evaporation are expected,
particularly in the warm seasons (summer and spring). It is important to note that the
seasonal distribution of precipitation has shown by the models to slightly change in the future
scenarios as compared to the baseline distribution, with increases in precipitation in summer
and decreases in evaporation in the other seasons under climate change scenarios. Total
annual precipitation, however, is predicted not to change in the SEQ region. This
demonstrates that with a growing economy and population, and considering that annual
rainfall will remain unchanged, SEQ will have to either increase water availability through
construction of additional reservoirs or desalination plants, or decrease the losses of water
through, for instance, the employment of technologies to reduce evaporation from the
existing reservoirs.
28
As a result of a changing climate, particularly higher surface air temperatures, lake
temperature will also be different. The average volume-average temperature of Wivenhoe
Dam is likely to change from 19oC in the baseline scenario to 19.5oC and 20.5oC in the
periods of 2030-2050 and 2070-2090 respectively. The changes in lake temperature are
more pronounced in winter and spring than in the other seasons in 2030-2050. For 20702090, significant lake warming is expected in summer, winter and spring. The results also
showed that Wivenhoe Dam has been warming over the past decades and will warm at even
higher rates towards the end of the century.
In regard to the research methodology, the results of this study were derived from
thermodynamic modelling of Wivenhoe Dam using forcings obtained from nine GCMs
downscaled over the SEQ region. The nine GCMs were selected by the CSIRO, Australia,
from the 23 available GCMs, based on their efficacy in simulating the control climate of SEQ.
The results of the simulations from this study were presented and assessed in an ensemble
projection of the nine models in order to account for the uncertainties implicit in the choice of
the GCMs. Hence, the results were shown as a range of probable outcomes or, more
precisely, as an envelope of projections forced by the nine GCMs. This multi-model
approach has been found to provide better forecasts than single-model approaches, in terms
of skill, reliability and consistency, particularly when considering many variables or
diagnostics (Johnson & Sharma, 2009). Attempts to reduce the biases in future projections
should never be made by removing models arbitrarily from an analysis. Although the current
approach recognises the impossibility of obtaining an exact magnitude for future evaporation
and temperature changes of Wivenhoe Dam, it at least provides an initial insight into the
relevant trends, as well as into what is most likely to happen in the future.
The approach of this study has international applicability as the method used to estimate
evaporation differs from the traditional methods, mostly based on the Penman model. The
method in this study takes into account the temporal and spatial changes in lake water
29
temperature, driven by meteorological boundary conditions derived from the downscaled
climate change projections. This method is generic and can be applied to different regions
and not just to the case presented here. Moreover, we recommend that studies like this
should be incorporated in local long-term integrated water resources management plans to
account for the probable changes in water availability. This will then assist in sound decisionmaking concerning water resources management and planning. The adoption of this
methodology to other reservoirs in Australia in order to determine a national trend is what we
recommend for further investigation.
8. Acknowledgement
We acknowledge the modelling groups, the Program for Climate Model Diagnosis and
Intercomparison (PCMDI), and the WCRP's Working Group on Coupled Modelling (WGCM)
for their roles in making available the WCRP CMIP3 multi-model dataset. Support of this
dataset is provided by the Office of Science, U.S. Department of Energy. We also thank the
CSIRO and the Bureau of Meteorology for making available the downscaled data from the
nine models utilised in this study.
Funding for this project was provided by Griffith Climate Change Response Program, Griffith
University Postgraduate Research School through the GUPRS scholarship, the Australian
Government Department of Innovation, Industry, Science and Research through the IPRS
scholarship, the Griffith School of Engineering and the Urban Water Research Security
Alliance.
30
9. References
Brutsaert, W. 1982. Evaporation into the atmosphere, Dordrecht, Reidel. 299 p.
Charles, S.P., Bari, M.A., Kitsios, A., Bates, B.C. 2007. Effect of GCM bias on downscaled
precipitation and runoff projections for the Serpentine catchment, Western Australia. Int. J.
Climatol., 27 (12), pp. 1673-1690.
Chu, J.T., Xia, J., Xu, C., Singh, V.P.,2010. Statistical downscaling of daily mean
temperature, pan evaporation and precipitation for climate change scenarios in Haihe River,
China. Theor.Appl. Climatol., (1-2), pp. 149-161.
Collins, W.D. et al. 2006. The Community Climate System Model Version 3 (CCSM3). J.
Climate, 19 (11), pp. 2122-2143.
Craig, I., Green, A., Scobie, M., Schmidt, E. 2005. Controlling evaporation loss from water
storages. Report 1000580/1, National Centre for Engineering in Agriculture, Toowoomba.
207 p.
CSIRO and BOM. 2007. Climate change in Australia. Technical report of CSIRO, Bureau of
Meteorology and the Australian Greenhouse Office in partnership with the Australian Climate
Change Science Program, Canberra. 148 p.
Delworth, T.L. et al. 2006. GFDL's CM2 Global Coupled Climate Models. Part I: Formulation
and simulation characteristics. J. Climate, 19 (5), pp. 643-674.
Diansky, N.A., Volodin, E.M. 2002. Simulation of present-day climate with a coupled
atmosphere-ocean general circulation model. (Engl. Transl.). Izv. Atmos. Ocean. Phys., 38
(6), pp. 732-747.
Gordon, C. et al. 2000. The simulation of SST, sea ice extents and ocean heat transports in
a version of the Hadley Centre coupled model without flux adjustments. Climate Dyn., 16,
pp. 147-168.
31
Gordon, H.B. et al. 2010. The CSIRO Mk3.5 Climate Model. Technical Report no. 021,
Centre for Australian Weather and Climate Research (a partnership between CSIRO and the
Bureau of Meteorology), Melbourne. 62 p.
Gordon, H.B. et al. 2002. The CSIRO Mk3 Climate System Model. Technical Paper 60,
CSIRO Atmospheric Research, Melbourne. 130 p.
Haerter, J.O., Hagemann, S., Moseley, C., Piani, C. 2011. Climate model bias correction and
the role of timescales. Hydrol. Earth Syst. Sci., 15, pp. 1065-1079.
Hasumi, H. et al. 2004. K-1 coupled GCM (MIROC) description. Technical Report no. 1 of
the K-1 model developers, Centre for Climate System Research, University of Tokyo, Tokyo.
34 p.
Hay, L.E., Wilby, R.L., Leavesley, G.H. 2000. A comparison of delta change and downscaled
GCM scenarios for three mountainous basins in the United States. J. Am. Water Resour. As.
36, pp. 387-397.
Helfer, F., Zhang, H., Lemckert, C. 2011. Modelling of lake mixing induced by air-bubble
plumes and the effects on evaporation. J. Hydrol., 406, pp. 182-198.
Hocking, G.C., Sherman, B.S., Patterson, J.C. 1988. Algorithm for selective withdrawal from
stratified reservoir. J. Hydraul. Div. ASCE, 114, pp. 707-719.
Imberger, J., Patterson, J.C. 1981. A dynamic reservoir simulation model - DYRESM. In:
Fischer, H.B. (ed.), Transport models for inland and coastal waters. New York: Academic
Press. pp. 310-361.
Imberger, J., Patterson, J.C., Hebbert, B., Loh, I. 1978. Dynamics of reservoir of medium
size. J. Hydraul. Div. ASCE, 104 (HY5), pp. 725-743.
32
IPCC. 2000. Special report on emissions scenarios (SRES). Special Report of Working
Group III of the Intergovernmental Panel on Climate Change, Cambridge University Press,
Cambridge, 570 p.
IPCC. 2007. Climate change 2007: The physical science. Contribution of working group I to
the fourth assessment report of the IPCC (AR4). Cambridge University Press, Cambridge.
996 p.
Ivey, G.N., Patterson, J.C. 1984. A model of the vertical mixing in Lake Erie in summer.
Limnol. Oceanogr., 29, pp. 553-563.
Johns, T.C. et al. 2006. The new Hadley Centre climate model HadGEM1: Evaluation of
coupled simulations. J. Climate, 19 (7), pp. 1327-1353.
Johnson, F., Sharma, A. 2009. Measurement of GCM skills in predicting variables relevant
for hydroclimatological assessments. J. Climate, 22, pp. 4373–4382.
Johnson, F., Sharma, A. 2010. A comparison of Australian open water body evaporation
trends for current and future climates estimated from class a evaporation pans and general
circulation models. J. Hydrometeor., 11, pp. 105–121.
Kay, A.L., Davies, H.N. 2008. Calculating potential evaporation from climate model data: A
source of uncertainty for hydrological climate change impacts. J. Hydrol., 358 (3-4), pp. 221239.
Kay, A.L., Jones, R.G., Reynard, N.S. 2006. RCM rainfall for UK flood frequency estimation.
II. Climate change results. J. Hydrol., 318 (1-4), pp. 163-172.
Kirono, D.G.C., Kent, D.M. 2011. Assessment of rainfall and potential evaporation from
global climate models and its implications for Australian regional drought projection. Int. J.
Climatol., 31, pp. 1295-1308.
33
Leander, R., Buishand, T.A., van den Hurk, B.J.J.M., de Wit, M.J.M. 2008. Estimated
changes in flood quantiles of the river Meuse from resampling of regional climate model
output. J. Hydrol., 351 (3-4), pp. 331-343.
Lenderink, G., Buishand, A., van Deursen, W. 2007. Estimates of future discharges of the
river Rhine using two scenario methodologies: direct versus delta approach. Hydrol. Earth
Syst. Sci., 11, pp. 1145-1159.
Martin, G.M. et al. 2006. The physical properties of the atmosphere in the new Hadley
Centre Global Environmental Model, HadGEM1. Part 1: Model description and global
climatology. J. Climate, 19 (7), pp. 1274-1301.
McGregor, J.L. 2005. C-CAM: Geometric aspects and dynamical formulation. CSIRO
Atmospheric Research Technical Paper no. 70. 43 p.
McGregor, J.L., Dix, M.R. 2001. The CSIRO conformal-cubic atmospheric GCM. In:
Proceedings of the IUTAN Symposium on Advances in Mathematical Modelling of
Atmospheric and Ocean Dynamics. July 02 - 07, 2000, Dordrecht, pp. 197-202.
McGregor, J.L., Dix, M.R. 2008. An updated description of the Conformal-Cubic Atmospheric
Model. In: Hamilton, K. & Ohfuchi, W. (eds.). High resolution numerical modelling of the
atmosphere and ocean. Springer. pp. 51-76.
McGregor, J.L., Nguyen, K. 2010. Dynamical downscaling from climate change experiments.
Final report for Project 2.1.5b, CSIRO Land and Water. 21 p.
Meehl, G.A. et al. 2007. The WCRP CMIP3 multimodel dataset: A new era in climate change
research. B. Am. Meteorol. Soc., 88 (9), pp. 1383-1394.
Monika, W., Reuter, J.E., Schladow, S.G. 2009. Lake warming favours small-sized
planktonic diatom species. Proc. R. Soc. B., 276, pp. 427-435.
34
Nguyen, K., Katzfey, J., McGregor, J. 2011. Global 60 km simulations with CCAM:
evaluation over the tropics. Clim. Dynam., pp. 1-18, (accepted, doi 10.1007/s00382-0111197-8).
Nunez, M., McGregor, J.L. 2007. Modelling future water environments of Tasmania,
Australia. Climate Research: Interactions of Climate with Organisms, Ecosystems, and
Human Societies, 34, pp. 25-37.
Patterson, J.C., Hamblin, P.F., Imberger, J. 1984. Classification and dynamic simulation of
the vertical density structure of lakes. Limnol. Oceanogr., 29 (4), pp. 845-861.
Patterson, J.C., Imberger, J. 1989. Simulation of bubble plume destratification systems in
reservoirs. Aquat. Sci., 51, pp. 3-18.
Perkins, S.E., Pitman, A.J., Holbrook, N.J., McAneney, J. 2007. Evaluation of the AR4
climate models simulated daily maximum temperature, minimum temperature, and
precipitation over Australia using probability density functions. J. Climate, 20, pp. 4356-4376.
Pope, V., Gallani, M.L., Rowntree, P.R. and Stratton, R.A. 2000. The impact of new physical
parameterizations in the Hadley Centre climate model: HadAM3. Clim. Dynam., 16, pp. 123146.
Raupach, M.R. et al. 2007. Global and regional drivers of accelerating CO2 emissions. Proc.
Nat. Acad. Science, 104, pp. 10288-10293.
Ringer, M.A. et al. 2006. The physical properties of the atmosphere in the new Hadley
Centre Global Environmental Model, HadGEM1. Part 2: Aspects of variability and regional
climate. J. Climate, 19 (7), pp. 1302-1326.
Roeckner, E. et al. 2003. The atmospheric general circulation model ECHAM5. Part I: model
description. MPI report no. 349, Max Planck Institute for Meteorology, Hamburg. 127 p.
35
Sahoo, G.B., Schladow, S.G. 2008. Impacts of climate change on lakes and reservoirs
dynamics and restoration policies. Sustain. Sci., 3, pp. 189-199.
Sherman, F.S., Imberger, J., Corcos, G.M. 1978. Turbulence and mixing in stably stratified
waters. A. Rev. Fluid Mech., 10, pp. 267-288.
Smith, I., Chandler, E. 2010. Refining rainfall projections for the Murray Darling Basin of
south-east Australia - the effect of sampling model results based on performance. Climatic
Change, 102 (3), pp. 377-393.
Spigel, R.H., Imberger, J. 1980. The classification of mixed-layer dynamics in lakes of small
to medium size. J. Phys. Oceanogr., 10 (7), pp. 1104-1121.
Thatcher, M., McGregor, J. L. 2011. A technique for dynamically downscaling daily-averaged
GCM datasets using the Conformal Cubic Atmospheric Model. Mon. Weather Rev., 139, pp.
79-79-95.
van Roosmalen, L., Christensen, J.H., Butts, M.B., Jensen, K.H., Refsgaard, J.C. 2010. An
intercomparison of regional climate model data for hydrological impact studies in Denmark.
J. Hydrol., 380, pp. 406-419.
Watterson, I.G. 1996. Non-dimensional measures of climate model performance. Int. J.
Climatol., 16, pp. 379-391.
Xu, Y., Yang, Z.L. 2012. A method to study the impact of climate change on variability of
river flow: an example from the Guadalupe River in Texas. Climatic Change (in press).
Yao, X. 2008. An improved 2-D numeric model for open-water daily evaporation estimation.
BA of Science Thesis, Griffith University, Gold Coast, Australia. 59 p.
Yeates, P.S., Imberger, J. 2003. Pseudo two-dimensional simulations of internal and
boundary fluxes in stratified lakes and reservoirs. Intl. J. River Basin Management, 1 (4), pp.
297-319.
36
Yilmaz, A.G. and Imteaz, M.A. 2011. Impact of climate change on runoff in the upper part of
the Euphrates basin. Hydrolog. Sci. J., 56 (7), pp. 1265-1279.
37